Factor analysis- correlations and estimation method

General help and assistance with jamovi. Bug reports can be made at our issues page: https://github.com/jamovi/jamovi/issues . (If you're unsure feel free to discuss it here)
Post Reply
OliviaKirtley
Posts: 1
Joined: Wed Jul 25, 2018 8:35 am

Factor analysis- correlations and estimation method

Post by OliviaKirtley »

Hi there,

I was wondering if JAMOVI automatically uses polychoric correlations for ordinal data in CFA and EFA?

Also, what is the estimation method used for CFA?

Thanks in advance for your help!

Olivia
User avatar
Ravi
Posts: 194
Joined: Sat Jan 28, 2017 11:18 am

Re: Factor analysis- correlations and estimation method

Post by Ravi »

Hi Olivia,

At the moment we don't use polychoric correlations, we assume that everything is continuous.

For CFA we use the maximum likelihood, the default estimator in lavaan (which is the R package that the CFA is based on). See, http://lavaan.ugent.be/tutorial/est.html

Cheers,
Ravi
Kbadar
Posts: 1
Joined: Mon Jul 19, 2021 3:43 am

Re: Factor analysis- correlations and estimation method

Post by Kbadar »

Hi,

Got to see this very old forum. The same question as Olivia had put earlier. Is there any possibility that we see a possibility of factor analysis with polychoric correlations in Jamovi anytime soon? Further, it will highly be appreciated if the following other estimates for EFA/CFA from the laavan package are introduced:

"GLS": generalized least squares. "WLS": weighted least squares (sometimes called ADF estimation).
"DWLS": diagonally weighted least squares
"ULS": unweighted least squares
"DLS": distributionally-weighted least squares
"PML": pairwise maximum likelihood

Regards
User avatar
Ravi
Posts: 194
Joined: Sat Jan 28, 2017 11:18 am

Re: Factor analysis- correlations and estimation method

Post by Ravi »

Hi,

So at the moment CFA can indeed only work with continuous variables (and when you add an ordinal variable, it's automatically transformed to a continuous one). For estimation method, we use FIML cause it can deal with missing values well. However, FIML does not work with non-continuous variables. So to make this work, the choice of estimation method suddenly needs to be conditional on whether a categorical variable is added, making the analysis more complicated than I intended. It's always a tight balance between an analysis being easy to use, and still as complete as possible. For the CFA it was mostly my goal to give people a basic and simple analysis that can also easily be taught to students. For more complex analysis, it would be better if there were a dedicated SEM module that aims to provide everything the researcher needs. I know that there's a group of people, that are working on such a dedicated SEM module, and I think a more complex CFA (that includes for example group comparisons, second order factor, and polychoric correlations) would be a great addition to such a module.

Cheers,
Ravi
Kamil Janowicz
Posts: 2
Joined: Wed Dec 20, 2023 6:01 pm

Re: Factor analysis- correlations and estimation method

Post by Kamil Janowicz »

Hi.I have the same question. For now, what estimation is using the CFA in JAMOVI?
Post Reply